SCHOOL OF TECHNOLOGY
CHEMICAL ENGINEERING GRADUATE PROGRAM
Gabriela Xavier de Oliveira
CFD AND OPTICAL ANALYSIS OF A LUMINESCENT SOLAR CONCENTRATOR-BASED PHOTOMICROREACTOR
FLORIANÓPOLIS 2019
Gabriela Xavier de Oliveira
CFD AND OPTICAL ANALYSIS OF A LUMINESCENT SOLAR CONCENTRATOR-BASED PHOTOMICROREACTOR
Master thesis for the degree of Master in Chemical Engineering presented to the Graduate Program in Chemical Engineering at the Federal University of Santa Catarina.
Advisor: Prof. Dr. Cíntia Soares Co-advisor: Prof. Dr. Natan Padoin Prof. Dr. Timothy Noël
Florianópolis 2019
Ficha de identificação da obra
A ficha de identificação é elaborada pelo próprio autor. Orientações em:
Gabriela Xavier de Oliveira
CFD AND OPTICAL ANALYSIS OF A LUMINESCENT SOLAR CONCENTRATOR-BASED PHOTOMICROREACTOR
O presente trabalho em nível de mestrado foi avaliado e aprovado por banca examinadora composta pelos seguintes membros:
Prof.ª Claudia Sayer, Dr.ª
Universidade Federal de Santa Catarina
Prof. Bruno Augusto Mattar Carciofi, Dr. Universidade Federal de Santa Catarina
Certificamos que esta é a versão original e final do trabalho de conclusão que foi julgado adequado para obtenção do título de mestre em Engenharia Química pelo Programa de Pós-Graduação em Engenharia Química da Universidade Federal de Santa Catarina.
____________________________ Prof.ª Dr.ª Cíntia Soares Coordenadora do Programa
____________________________ Prof.ª Dr.ª Cíntia Soares
Orientadora
Florianópolis, 07 de junho de 2019.
Cintia
Soares:020
40162909
Digitally signed by Cintia Soares:02040162909 DN: cn=Cintia
Soares:02040162909, o=ICPEdu, ou=UFSC - Universidade Federal de Santa Catarina
Date: 2019.08.20 12:53:13 -03'00'
Cintia
Soares:020
40162909
Digitally signed by Cintia Soares:02040162909 DN: cn=Cintia
Soares:02040162909, o=ICPEdu, ou=UFSC - Universidade Federal de Santa Catarina
This work is dedicated to my sources of strength, my parents and my grandmother.
ACKNOWLEDGEMENTS
First of all, I would like to express my gratefulness for being able to take a high-quality graduate course with highly qualified professors. Science in Brazil is a reality.
I thank my parents, whom I love immensely, without them none of this would be happening.
To my family, especially my brother, godmothers, and cousins. Thank you all for the affection and concern.
To my advisor, Prof. Cíntia Soares. Thank you for all the patience, support and commitment as professor and advisor. I am very proud to have been guided by you.
To my co-advisor, Prof. Natan Padoin. Thank you for your continued availability, your help was indispensable for this work.
Thanks to Prof. Timothy Noël and Dr. Dario Cambié for their attention and the information provided.
I also would like to thank my life friends, Priscila and Tairine. I know that I can always count on you both and I am very happy that we are always available to each other.
I also thank the great friends that I made in the masters, Ana Paula, Bernardo and Jessica, you guys are incredible.
A special thanks to all the members of the LabMAC group for all the happy moments I had in this lab.
Finally, I am thankful to CAPES (Coordination for the Improvement of Higher Education Personnel) and CNPq (National Council for Scientific and Technological Development) for financial support.
“…not all those who wander are lost...”
RESUMO
A sustentabilidade nos processos é uma questão que vem sendo alvo de muita discussão ao longo dos anos. Uma nova tecnologia que emprega esse processo limpo é o foto microrreator baseado na tecnologia de concentrador solar luminescente (LSC-PM), o qual utiliza a luz do sol para promover reações químicas. Neste estudo, um modelo em CFD foi desenvolvido para a simular a conversão de 9,10-difenilantraceno (DPA) no LSC-PM. Os resultados simulados foram comparados com os dados experimentais de conversão de DPA em função do tempo para diferentes potências, com boa concordância. Três funções (retangular, triangular e sinusoidal) foram criadas para representar a variação temporal da potência. Um sistema de controle feedforward foi implementado no sistema de reação para manter uma conversão de DPA estável mesmo com as variações de irradiação ao longo do tempo. O estudo do tempo de atraso no sistema de controle foi feito com quatro tempos de atraso, 0,5, 1, 5 e 10 s, onde foi verificado que o atraso de 0,5 s resulta em controle em tempo real e é perfeitamente realizável. Além disso, duas configurações geométricas diferentes foram construídas para avaliar a influência desse parâmetro sobre o desempenho do sistema. Simulações da distribuição de fótons no LSC-PM foram conduzidas com base no método de Monte Carlo, utilizando a luz do sol como fonte. O livre caminho médio das partículas luminescentes foi variado para encontrar a melhor resposta e a interferência do mesmo no sistema óptico. Ao analisar a potência que atinge a superfície do dispositivo e a potência de saída, um livre caminho médio de 1 mm foi considerado razoável e posteriormente foi adotado para todas as simulações. Este estudo permitiu observar a variação das perdas pela diminuição/aumento da concentração de partículas de corante. Uma alta concentração implica em auto-perdas, e uma concentração muito baixa em um aumento de transmissões. A potência que atinge o dispositivo pôde ser medida, sendo que 1,7771W foi medido no interior do sistema, enquanto 0,0183W foram coletados pela borda de saída. As perdas de transmissão puderam ser elucidadas através das medidas coletadas por uma placa fina instalada sob o foto microrreator, e provou que 0,058 W da potência perdeu-se por transmissão. Duas novas espessuras do fotomicroreator foram desenvolvidas, a fim de comprovar que a geometria interfere no comportamento óptico do sistema, possibilitando a realização de estudos adicionais futuramente visando melhorias no sistema. Baseado nisso, as duas novas geometrias mostraram que, ao diminuir a espessura do dispositivo, as perdas são ampliadas no sistema.
RESUMO EXPANDIDO
Introdução
A utilização desenfreada de combustiveis fósseis, a qual gera condições climáticas desfavoráveis em nível global, gera a necessidade de novas rotas energéticas. Dentre os recursos energéticos renováveis, a luz solar mostra-se uma excelente alternativa, uma vez que é abundante em vários lugares do mundo, diferentemente dos combustíveis fósseis, que são escassos. Um dos usos da luz solar para a geração de energia é o concentrador solar luminescente (LSC), tecnologia desenvolvida há mais de trinta anos atrás, durante a crise energética, e ainda amplamente empregada. O sistema funciona de um modo simples, onde os fótons, após penetrarem na superficie superior do dispositivo são presos no mesmo e reemitidos para as bordas da placa onde celulas fotovoltaicas são alocadas e resposáveis pela conversão em energia eletrica. Porém, não é apenas a industria energética que necessita de rotas mais sustentáveis. Estimulados pelo aumento dos processos limpos de geração de energia, os processos sustentáveis na industria química começaram a ganhar espaço. Os processos fotoquímicos, os quais em seus primórdios eram efetuados com luz solar, voltaram a repensar e reutilizar o uso da mesma. Com isso em mente, Cambié et al. (2017), propuseram um foto microrreator para promover reações químicas a partir de luz solar, o fotomicrorreator baseado na tecnologia de concentrador solar luminescente (LSC-PM). Tal foto microrreator constitiu-se em uma fusão sinérgica entre o LSC e microcanais, gerando um microrreator de baixo custo e simples, capaz de realizar sinteses químicas, o qual leva como reação de referência a cicloadição [4+2] de 9,10 difenilantraceno (DPA). Dentre as várias questões presentes quando se trabalha com luz solar, mudanças na irradiação solar durante o dia tornam-se um dos maiores desafios. Com isto em mente, Zhao, et al. (2018), implementaram um sistema de controle feedforward no LSC-PM, o qual foi responsável por manter a conversão constante independente das oscilações de irradiação durante o dia. Porém, apesar das variadas vantagens do LSC-PM, algumas desvantagens ainda estão presentes no sistema. Além das mudança da irradiação durante o dia já citadas, as perdas de energia do sistema e o fluxo de fótons atingindo os microcanais podem ser alguns dos fatores que interferem na eficiência do sistema. Simulações de traçados de raios e em CFD podem se tornar ferramentas capazes de elucidar e minimizar essas questões. Objetivos
Tendo em vista que sistemas fotoquímicos apresentam perdas energéticas, simulações de traçados de raios podem ser realizadas nos sistemas. Deste modo, sugeriu-se a simulaçao de traçado de raios para quantificar as perdas energéticas do sistema e minimizá-las. Este trabalho visa também, implementar um modelo CFD do LSC-PM, simulando o sistema de controle visando primeiramente uma validação, para posterior otimização do sistema. Além de, avaliar a performance do LSC-PM com e sem o controle, variados tempos de atraso e com diferentes intensidades de luz. Novas geometrias foram propostas também para otimização da performance do foto microrreator.
Metodologia
A metodologia, assim como os resultados, dividiu-se em duas seções: (i) simulação em CFD; (ii) simulação de traçado de raios. O primeiro item, (i) simulação em CFD, foi efetuado no software COMSOL Multiphysics® (versão 5.3a). A metodologia deste item (i) seguiu as seguintes etapas: a) desenvolvimento da geometria do LSC-PM no COMSOL Multiphysics® com as dimensões originais; b) implementação do teste de independência de malha proposto por Celik et al. (2008).; c) determinação da constante de velocidade de pseudo-primeira ordem, através de dados fornecidos pelo Noël Research Group; d) após a validação do modelo computacional com o experimental, variações da intensidade de luz foram representadas através de distúrbios no software COMSOL Multiphysics® na forma sinusoidal, retangular e triangular; e) desenvolvimento do sistema feedforward de controle para manter a conversão constante independente da irradiação do dia. Para tal, foi necessário realizar um balanço de massa a fim de descobrir como a velocidade varia com a potência, utilizando como hipóteses: estado estacionário e plug-flow na direção x. f) variação do tempo de atraso a fim de verificar o comportamento do sistema; g) desenvolvimento de duas novas geometrias para minimizar a formação de zonas mortas no LSC-PM através da função fillet e elipse do COMSOL Multiphysics®. Quanto ao segundo item, (ii) simulação de traçado de raios, a metodologia deste tópico seguiu as seguintes etapas: a) implementar a geometria original do LSC-PM juntamente com as características ópticas no software LightTools®, definindo-se nesse caso, 100% absorção nas laterais, e “tipo Fresnel” no topo e base do reator; b) o próximo passo da metodologia tratou-se então do procedimento de dopagem de corante do sistema, onde o espectro de absorção e emissão (WIELAND, 2016), e o rendimento quântico (BASF, 1997) foram definidos no sistema; c) o livre caminho médio (MFP) devido à ausência de dados, foi assumido, utilizando um caso base citado por Chavéz et al. (2017); d) na modelagem da fonte de luz, considerou-se a data e as coordenadas originais da região onde os experimentos foram realizados, obtidos através dos dados suplementares de Zhao et al. (2018); e) após a inserção de todos os parâmetros, uma variação na espessura do foto microrreator foi feita de modo a observar a interferência da mesma na performance do sistema; f) receptores foram acoplados em três superfícies do foto microrreator para quantificar a potência atingindo o mesmo.
Resultados e Discussão
Na (i) simulação em CFD: os dados simulados apresentaram uma excelente concordância com os dados experimentais com um erro máximo <13%. Deste modo, pode ser estabelecido que o modelo simulado representa satisfatoriamente a conversão de DPA em diferentes intensidades de luz. Quando se diz respeito ao sistema do controle, observou-se que o sistema sem nenhum controle, não apresentou uma conversão constante, como era esperado. Pelo contrário, quando a irradiação diminuía, a conversão diminuía de mesmo modo (de~ 80% para < 10%). Ao aplicar o sistema de controle o comportamento oposto foi observado, onde o sistema de controle foi capaz de manter a conversão em torno de ~90% mesmo com as diferentes intensidades de luz. Quanto as três funções desenvolvidas para representar a intensidade de luz (triangular, sinusoidal e retangular) com diferentes tempos de atraso: 0,5, 1, 5 e 10 s, observou-se que a diferença de conversão entre os tempos de atraso de 0,1 e 1 é praticamente nulo. Diferentemente, o tempo de
atraso de 5 s demonstrou diferenças substanciais na conversão. Consequentemente, o tempo de atraso de 10 s foi o que mais demonstrou desvio do set-point. Dentre os distúrbios, a função triangular foi a que mais demonstrou um decaimento drástico com a diminuição da intensidade da luz. Comportamento que pode ser explicado devido ao fato de que quando ocorre uma diminuição repentina da intensidade da luz, sem uma diminuição imediata da vazão, o resultado é uma diminuição da taxa de reação e consequentemente na conversão. Com a mudança das geometrias, observou-se que com a geometria elipse obteve-se a maior conversão 91,27%, quando comparada a conversão da geometria original, 90,67%. A geometria fillet resultou em uma conversão de 90,96%, não tão diferente do desempenho obtido com a geometria original. Deste modo, a geometria original, mesmo com suas extremidades retangulares, apresentou comportamento aceitável quando comparado as duas novas geometrias, não demonstrando necessidade de substituição. Na (ii) simulação de traçado de raios: algumas perdas foram elucidadas, onde, 1,771 W foram coletados incidindo a superfície do reator, e apenas, 0,117 W foram absorvidos pelos microcanais. Sendo que, 0,058 W foram perdidos por transmissão dos 0,117 W totais. A potência de saída também foi coletada visando uma futura validação com os dados de Zhao et al. (2018), onde encontrou-se um valor de 0,00183 W atingindo as laterais (saída), ou seja, apenas uma porcentagem de 7,39% dos raios acaba por atingir as laterais do reator (considerando 100% de absorção). O livre caminho médio (MFP) foi então avaliado, o qual era desconhecido. Através de intervalos que variavam de 0,1−20mm, encontrou-se o valor ideal de MFP≈1mm. Onde este valor foi escolhido devido ao fato de que com valores muito altos de MFP observou-se que ocorria diminuições na potência absorvida e na potência de saída do reator, o que é causado pelo alto índice de transmissão. E valores muito baixos de MFP geravam também concentrações muito altas de corante no sistema, ocasionando muitas perdas por auto-absorção. A modificação da espessura da geometria mostrou um resultado promissor para futuros estudos, representando que tal fator influencia na eficiência do sistema tanto na fluidodinâmica quanto na performance óptica. Conforme aumentou-se a espessura do reator, houve também, de mesmo modo, um leve aumento na potência absorvida e na potência de saída.
Considerações Finais
Este estudo destacou que a ferramenta de CFD pode ser aplicada com sucesso na investigação da fluidodinâmica em um microrreator. O modelo de CFD desenvolvido mostrou boa concordância com os dados experimentais, possibilitando a validação do modelo. O sistema feedforward implementado no sistema de reação foi capaz de manter a conversão no valor alvo apesar das mudanças da irradiação de luz. Além disso, a geometria original mostrou bom comportamento quando a conversão foi estudada, não havendo necessidade de alteração. A partir da simulação óptica, foi possível elucidar a potência que atinge o sistema. A potência que atinge a parte superior do dispositivo e a potência de saída também puderam ser coletadas. A investigação das perdas provou que estudos a respeito da otimização ainda podem ser feitos para melhorar o desempenho do sistema (e.g. otimização da geometria). A validação é o próximo passo principal, tendo em vista que, para fazer a otimização, as simulações devem ter um bom acordo com os dados experimentais.
ABSTRACT
Process sustainability is a question that has attracted significant attention in the last decades. A novel technology that employs this clean process is the luminescent solar concentrator-based photomicroreactor (LSC-PM), a microreactor that utilizes sunlight to drive chemical reactions. In this study, a CFD model was built for the simulation of the conversion of 9,10-diphenylanthracene (DPA) in the LSC-PM. The numerical results were compared with the experimental data of DPA conversion versus time for different powers, and the agreement was quite satisfactory. Three functions (rectangular, triangular and sinusoidal) were created in order to represent the temporal variation of the power. A feedforward control system was implemented in the reaction system in order to maintain a stable DPA conversion due to power variations over time. The study of the time delay in the control system was made with four levels, 0.5, 1.5 and 10 s, where it was verified that, in fact, a 0.5 s delay resulted in a realizable real-time control. Furthermore, the influence of different geometrical configurations for the microchannels on the reactive flow was investigated. Moreover, optical simulations were carried out to quantify the photon distribution in the system based on Monte Carlo method. These ray-tracing simulations used sunlight as a source. The mean free path (MFP) of the luminescent particles was variated in order to find the best response and the interference of this parameter on the performance of the optical system. When analyzing the power that reaches the device and the output power, a 1 mm MFP was considered adequate and adopted for all the simulations. This study allowed the observation of the variation of losses by the decrease/increase of the concentration of dye particles. A high concentration implies in self-losses, while a very low concentration implies in higher transmissions. The power attaching the device by sunlight irradiation could be measured, where 1.7771 W was found impinging in the system, while 0.0183 W was collected by the output edge. The transmission losses could be elucidated by attaching a thin plate under the photomicroreactor, and it proved that 0.058 W of the power was lost by transmission. Two new thickness of the photomicroreactor were evaluated, in order to prove that it interferes in the optical behavior of the system, allowing further improvement studies on the system. These two new dimensions showed that by decreasing the thickness of the device the losses are consequently increased.
LIST OF FIGURES
Figure 1 - Simplified Jablonski diagram. ... 25
Figure 2 - Quenching cycles of a photocatalyst. ... 28
Figure 3 - Schematic representation of the LSC-PM system. ... 32
Figure 4 - Representation of the LSC system... 33
Figure 5 - Representation of possible fates of the photons in the moment that the light strikes the LSC-PM device. ... 36
Figure 6 - 9,10 diphenylanthracene conversion to its respective endoperoxide. ... 40
Figure 7 - Representation of the LSC-PM real-time reaction control system. ... 42
Figure 8 – Flow scheme of the indoor kinect experimental procedure. ... 45
Figure 9 – (1) Representation of the LSC-PM built in COMSOL Multiphysics® (version 5.3a) software. (2) Dimensions of the LSC-PM device. ... 46
Figure 10 - Relation of the rate constant (k) and the power (P). ... 52
Figure 11 - Different forms for power disturbances as a function of time developed in the COMSOL Multiphysics® software: (a) rectangular, (b) triangular and (c) sinusoidal. ... 53
Figure 12 - Block diagram of the LSC-PM simulation. ... 55
Figure 13 - Velocity field at one of the microchannel’s edges... 56
Figure 14 - Logical flow chart of the ray-tracing simulation. ... 57
Figure 15 - Representation of the LSC-PM geometry imported in the software LightTools® (Synopsys®, Inc.). ... 58
Figure 16 - Absorption and emission spectrum of Red dye 305 ... 62
Figure 17 - Comparison between experimental data and those obtained by the numerical simulations. ... 66
Figure 18 - Comparison of LSC-PM reactor with and without conversion control, considering 90% as target conversion. ... 68
Figure 19 - Comparative evaluation for different time delays for power format: (a) rectangular, (b) triangular and (c) sinusoidal. ... 71
Figure 20 - Velocity field in three different geometries: (a) original, (b) slightly curved and (c) highly curved. ... 73
Figure 21- Ray-tracing simulations in the LightTools® software*: (a) Ray-tracing simulation on the PM, (b) Ray-tracing simulation with a thin plate under the LSC-PM. ... 74 Figure 22 - MFP influence on the energetic efficiency of the LSC-PM. ... 76 Figure 23 - Influence of the waveguide thickness on the system’s power distribution. ... 77
LIST OF TABLES
Table 1- Physical properties of the species. ... 50 Table 2 - Extinction coefficient for the PDMS on different wavelengths. ... 60 Table 3 - Emission and absorption spectrum of Red 305 dye at different wavelengths. ... 63 Table 4 - Error performance for different time delays in the LSC-PM control system. ... 72
LIST OF ABBREVIATIONS CFD Compututional Fluid Dynamics
DPA Diphenylanthracene
ESA GCI IAE
Excited State Absorption Grid Convergence Index Integral Absolute Error
IC Internal Conversion
ISC ISE ITAE LEDs
Internal System Crossing Integral Square Error
Integral of time multiplied absolute error Light-Emitting Diodes
LSC Luminescent Solar Concentrator
LSC-PM Luminescent Solar Concentrator Photomicroreactor MB
MFP
Methylene Blue Mean Free Path
PDMS Polydimethylsiloxane
PFA Perfluoroalkoxyalkane
PMMA Polymethylmethacrylate
PV Photovoltaic
SET Single Electron Transfer TIR Total Internal Reflection
LIST OF SYMBOLS
υx, υy, υz velocity components (m ∙ s−1)
Ci concentration of specie i (mol ∙ m−3) Ci,0 initial concentration of specie i (mol ∙ m−3)
k rate constant (s−1)
Fi,0 molar feed rate (mol ∙ s−1)
X conversion (%) V volume of reactor (m3) P power (W) τ space time (s) Q̇ volumetric rate (m3∙ s−1) A area (m2) ρ density ( kg ∙ m−3) P pressure (Pa)
τij component of stress tensor
μi dynamic viscosity for each species i in the mixture (Pa ∙ s)
xi mole fraction of specie i
n
⃗ normal velocity
t tangential velocity
Ji (x,y,z) molecular diffusion flux of the species in x,y,z direction (mol ∙ m−2∙ s−1)
Di,m molecular diffusion coefficient for specie i in the mixture (m2∙ s−1)
Dij molecular diffusion coefficient for the pair i-j (m2∙ s−1)
ϕj association parameters
T absolute temperature (K)
Mw,j molecular weight (g ∙ mol−1)
1O2 singlet oxygen
α absorption coefficient 𝜆 wavelength (nm) α(𝜆) absorption spectrum 𝑘 extinction coefficient n refractive index
CONTENTS 1 INTRODUCTION ... 21 1.1 MOTIVATION ... 21 1.2 OBJECTIVES ... 23 1.2.1 General objective ... 23 1.2.2 Specific objectives ... 23 2 THEORETICAL BACKGROUND ... 24 2.1 SUNLIGHT ... 24 2.2 PHOTOCHEMISTRY ... 24 2.2.1 Photoredox Catalysis ... 28 2.2.2 Photochemistry in microreactors ... 29
2.3 LUMINESCENT SOLAR CONCENTRATOR-BASED PHOTOMICROREACTOR (LSC-PM) ... 31
2.3.1 LSC principle ... 31
2.3.2 Host material ... 34
2.3.3 Luminescent particles ... 35
2.3.3.1 Light transport and loss factor ... 35
2.3.3.1.1 Ray-tracing simulations ... 37
2.3.4 Flow characterization ... 39
2.3.4.1 Control system ... 41
2.3.5 CFD applied to flow distribution characterization ... 43
3 MATERIALS AND METHODS ... 45
3.1 CFD SIMULATION ... 45
3.1.1 Geometry ... 46
3.1.2 Grid independence test ... 46
3.1.3 Mathematical model ... 47
3.2 RAY-TRACING SIMULATION ... 56 3.2.1 Geometry ... 57 3.2.1.1 Optical properties... 58 3.2.2 Waveguide ... 58 3.2.3 Luminescent particles ... 60 3.2.3.1 MFP ... 60
3.2.3.2 Absorption and emission spectra ... 61
3.2.3.3 Quantum yield ... 62
3.2.4 Source modeling ... 63
3.2.4.1 Receivers ... 64
3.2.5 Ray-tracing simulation input ... 64
4 RESULTS AND DISCUSSION ... 65
4.1 CFD SIMULATION ... 65
4.2 RAY-TRACING SIMULATION ... 74
5 CONCLUSIONS ... 79
6 RECOMMENDATIONS FOR FUTURE WORK ... 80
REFERENCES ... 81
APPENDIX A – Supplementary material for the CFD simulation ... 91
APPENDIX B – Supplementary material for ray-tracing simulation ... 95
1 INTRODUCTION 1.1 MOTIVATION
Among renewable energy resources, solar light is an excellent alternative, given that it is eco-friendly, free and abundant in several places around the world, unlike fossil fuels that are finite (KABIR et al., 2018; LEELADHAR; RATURI; SINGH, 2018; SANSANIWAL; SHARMA; MATHUR, 2018; YADAV et al., 2018). The use of solar energy requires capture, solar conversion and storage, and currently, is widely used for the generation of heat and electricity (LEWIS; NOCERA, 2006; SCHOLES et al., 2011).
About more than 30 years ago, during the energy crisis, a technology was proposed aiming at concentrating solar energy for electricity production, the so-called luminescent solar concentrators (LSCs) (DEBIJE, 2010; MEINARDI; BRUNI; BROVELLI, 2017; MORAITIS; SCHROPP; VAN SARK, 2018). The LSCs operate through the light penetration on the upper surface of the device where re-emission of absorbed photons occurs. Afterward, this light is concentrated along the edges of the plate where it is collected by small photovoltaic cells responsible for converting it in electrical energy (DEBIJE; VERBUNT, 2012; MORAITIS; SCHROPP; VAN SARK, 2018; SARK et al., 2008).
Recently, solar energy has also been used for the synthesis of chemical compounds, since the introduction of sunlight-driven photocatalysis (OELGEMÖLLER, 2016; SCHOLES et al., 2011; SCHULTZ; YOON, 2014). Inspired by the LSC technology, a novel luminescent solar concentrator-based photomicroreactor (LSC-PM) was developed at the Noël research group (http://www.noelresearchgroup.com/), Eindhoven University of Technology, The Netherlands, to promote fast chemistry under sunlight irradiation (CAMBIÉ et al., 2017a; ZHAO et al., 2018b). This microreactor is composed by a synergistic combination of luminescent solar concentrators and microchannels, allowing simple and inexpensive reactors, which
enables the efficient use of sunlight for photochemical transformations (CAMBIÉ et al., 2017b).
In addition, the use of microreactors has unique advantages for photochemical reactions, since it merges the small dimensions of the devices with the continuous flow mode (OELGEMÖLLER, 2012). Continuous flow operation presents a significant reduction in solvent requirements, ensuring sufficient light penetration (ODIBA et al., 2016), under totally customizable designs. Moreover, this operation mode is preferred for the synthesis of added-value compounds such those used in the pharmaceutical industry as active agents or intermediates given the excellent controllability, ensuring high quality for the chemicals synthesized.
However, despite all the reported advantages of the LSC-PM, there is still plenty of room for improvements, especially when one considers the modeling of photon distribution in the device and its coupling with the reactive flow taking place in the microchannels. The leakage of the photons when the irradiation impinges in the domain is a clear example. In this context, ray-tracing methods have been used as an effective tool to quantify photon losses and optimize the system, allowing to track the photon fate taking into account the effect of the optical properties of surfaces, materials and emission sources (KERROUCHE et al., 2014). This information is of paramount importance to minimize energy losses and intensify the photon flux reaching the microchannels, resulting in enhanced reaction rates.
Additionally, when sunlight is considered, one has to figure out the effect of fluctuations in the irradiation along the day, which impose a real challenge for this technology. An efficient alternative to overcome this problem has been recently reported, based on real-time feedforward control of the system, maintaining a constant conversion regardless of the light intensity that reaches the device during the day (ZHAO et al., 2018). CFD techniques can provide important insights about the behavior of microreactors, allowing to investigate control strategies as well as the effect of different geometrical configurations and operating conditions on the device’s performance (ODIBA et al., 2016), highlighting flow non-idealities for instance. Thus, once validated, a CFD model is a powerful tool for the optimization of photomicroreactors seeking for effective scale-out of the system.
Given this context, this work aims to implement a CFD model of a LSC-PM device applied to the cycloaddition of 9,10-diphenylanthracene (DPA), took as benchmark reaction (CAMBIÉ et al., 2017b), and numerically investigate a feedforward control strategy. In addition, it also intends to quantify the energy losses of the device exposed to sunlight irradiation through ray-tracing simulations.
1.2 OBJECTIVES
1.2.1 General objective
The general objective of this work is to investigate the 9,10-diphenylanthracene (DPA) synthesis in an LSC-PM under fluctuating solar irradiation by CFD, subjected to a feedforward control strategy, and implement a ray-tracing model to quantify energy losses in the device.
1.2.2 Specific objectives
To meet the general objective presented at item 1.2.1, this work is based on the following specific goals:
• obtain a light-dependent kinetic model for the cycloaddition of 9,10-diphenylanthracene (DPA) in an LSC-PM.
• validate a CFD model with experimental data available in the literature. • evaluate the performance of the LSC-PM under different light intensities. • compare the efficiency of the LSC-PM with/without a control system. • observe the response of the control system with different time delays. • propose new geometries for the improvement of the LSC-PM’s
performance.
• quantify the energies losses in the microreactor.
2 THEORETICAL BACKGROUND 2.1 SUNLIGHT
Clean energy is an environmental question that is gaining more space over the years. Considering that the release of carbon dioxide and greenhouse gases is deeply harmful to the environment and that the depletion of the oil reserves is at alarming levels, the quest for alternative energy sources is a central point for future technological development (OBAMA, 2017; SANSANIWAL; SHARMA; MATHUR, 2018). Among renewable energy resources, sunlight is an excellent alternative, once it is eco-friendly, free and abundantly available in several places around the world (KABIR et al., 2018; LEELADHAR; RATURI; SINGH, 2018; SANSANIWAL; SHARMA; MATHUR, 2018; YADAV et al., 2018).
Given the excellent potential of this energy source, several researches have explored ways to harvest sunlight not only for electricity generation but also to drive photocatalyzed chemical reactions. Pohlmann et al. (1997) demonstrated the suitability of sunlight-driven chemical reactions using moderately concentrated irradiance (OELGEMÖLLER; JUNG; MATTAY, 2007). Several studies have been devoted to developing sunlight-driven chemistry since then (CAMBIÉ et al., 2017a; OELGEMÖLLER, 2016; SCHOLES et al., 2011; SCHULTZ; YOON, 2014).
The use of solar light as a source of photons for photocatalyzed reactions is a promising field, including applications in degradation of liquid and gas-phase pollutants, synthesis of advanced materials and synthesis of chemicals, especially added-value molecules.
2.2 PHOTOCHEMISTRY
Photochemistry is an important technological alternative for the synthesis of organic compounds (CIANA; BOCHET, 2007). Commonly, the photochemical reactions are induced by high-energy ultraviolet (UV) or visible photons, which are responsible for the necessary energy to produce molecular transformations
(MUKAMEL, 1997; SHVYDKIV, 2012). Max Planck, the father of quantum physics, theorized that this energy was transferred in parts, or quanta, equal to ℎ𝑣, expressed according to Equation (1) (KAOARNOS, 1993).
𝐸 = 𝑛 ∙ ℎ𝑣 (1)
where 𝐸 is the energy of the photon, ℎ is Max Planck constant, 𝑣 is frequency, and 𝑛 is the number of photons.
Based on this theory, it is possible to say that molecules can exist in different electronic states, depending on the quantity of energy provided (WASSERBERG, 2006). These electronic states can be considered energetically unstable and chemically different from their corresponding ground states, with very brief life, around nanoseconds (RAZEGHIFARD, 2013). The possible transitions after the photoexcitation of a molecule are represented by the Jablonski diagram (Figure 1).
Figure 1 - Simplified Jablonski diagram.
These processes of transitions (Figure 1) can be explained as follows (WASSERBERG, 2006):
1. 𝑆𝑜 ground state: corresponds to the state of the molecule prior do the photon absorption.
2. 𝑆𝑜 + ℎ𝜈 → 𝑆1 singlet-singlet absorption: after the photoexcitation of the molecule, it is elevated from the ground state to a higher energy level, reaching the 𝑆1 state.
3. 𝑆𝑜 + 2 ℎ𝜈 → 𝑆1 two-photon singlet-singlet absorption: as the name indicates, two photons are absorbed by a molecule that is in the ground state, and almost simultaneously the molecule is elevated to the 𝑆1 state.
4. 𝑆𝑜 + ℎ𝜈 → 𝑇1 singlet-triplet absorption: occurs when a photon absorbed does not contain sufficient energy to excite the molecule to the 𝑆1 state; however, it is able to excite the molecule to the triplet state.
5. 𝑇1 + ℎ𝜈 → 𝑇𝑛 triplet-triplet absorption: occurs when a photon is absorbed by a molecule in 𝑇1 excited state, generating an excitation to an even higher triplet state 𝑇𝑛. It can be characterized as a form of excited-state absorption (ESA).
6. 𝑆1 → 𝑆𝑜+ ℎ𝜈 fluorescence: when a molecule in its excited singlet state decays radiatively between states of the same spin state. This can occur under the emission of a photon.
7. 𝑇1 → 𝑆𝑜+ ℎ𝜈 phosphorescence: the molecule located in the excited triplet state emits a photon and come back to the singlet ground stated.
8. 𝑆1 → 𝑆0+ ∆ internal conversion (IC): happens when a molecule that 𝑇𝑛 → 𝑇1+ ∆ is in an excited state goes through another electronic state of the same spin multiplicity. Then, the nonradiative decay due to the loss of vibrational quanta results in a release of thermal energy (∆).
9. 𝑆1 → 𝑇1+ ∆ intersystem crossing (ISC): when a molecule that is in 𝑇1 → 𝑆0 + ∆ an excited state goes through another electronic state with a different spin multiplicity. Then, the nonradiative decay due to the loss of vibrational quanta results in a release of thermal energy (∆).
In these processes, the shorter wavelengths of the photons allow direct interaction with molecular bonds. For this reason, the performance with UV is better due to its wavelength, 𝜆 < 387 nm (ASAHI, 2012; WIELAND, 2016). However, UV can generate undesired side reactions, with a negative impact on the selectivity of the reaction. On the other hand, when applying visible light these side reactions can be minimized. Moreover, the cost associated with the use of visible light is much lower than that of UV lamps, especially when large scale applications are considered. Therefore, visible light is an attractive alternative to overcome the aforementioned issues (ANGNES et al., 2015; WIELAND, 2016). However, a disadvantage of using sunlight to drive chemical transformations can be highlighted: the inability of organic molecules to absorb light in the visible range of the spectrum. Nevertheless, in order to absorb the visible light, a photocatalyst with color can be applied.
With the advances in photochemistry, several research topics have been developed. Among these, the most frequent topic related to visible light-driven chemical transformation is photoredox catalysis (WIELAND, 2016).
2.2.1 Photoredox Catalysis
The photoredox catalysis exploits the intensification of the redox activity in a photo-excited catalyst, which becomes a strong oxidant or a strong reductant. Considering that this activation mode occurs in moderate reaction conditions, e.g., with non-hazardous reagents and using visible light, it proves to be an interesting and promising application in pharmaceutical synthesis (ZELLER, 2016; NOËL, 2017).
As previously stated, one issue of using visible light is that organic molecules do not absorb in the visible region. Thus, in order to utilize this energy provided by the visible light, organic or transition-metal based photocatalysts with color can be applied. Those are able to harvest visible light, reaching their excited states and mediating the single-electron transfer (SET) processes from the excited state (NICHOLLS; LEONORI; BISSEMBER, 2016). According to Ghosh (2016), the process can follow two paths (Figure 2):
1. reductive quenching cycle: the excited photocatalyst receive an electron from a donor and then returns to the ground state in reduced form.
2. oxidative quenching cycle: the excited photocatalyst donate an electron to an acceptor and then returns to the ground state in oxidized form.
Figure 2 - Quenching cycles of a photocatalyst.
1- Reductive quenching cycle. 2 – Oxidative quenching cycle. Reference: Ghosh (2016).
The interaction among these electronically photo-excited catalysts and an organic molecule is able to generate reactive intermediates that can result in synthetically useful bond constructions (SKUBI; BLUM; YOON, 2016).
Several studies have been carried out using photoredox catalysis for the synthesis of compounds, and organometallic polypyridyl transition metal complexes and organic dyes can be considered the most applied and efficient photocatalysts in organic synthesis (KÇNIG et al., 2013; NARAYANAM; TUCKER; STEPHENSON, 2009; RAVELLI; FAGNONI; ALBINI, 2013; SCHULTZ; YOON, 2014; SHI; XIA, 2012; XUAN; XIAO, 2012). However, due the high cost and toxicity of transition-metal complexes, organic photosensitizers (metal-free) became an attractive option, since they commonly have low cost, long excited-state lifetime and high extinction coefficient (PITRE; MCTIERNAN; SCAIANO, 2016). The most common organic photosensitizers applied are methylene blue (CAMBIÉ et al., 2017a; KALAITZAKIS et al., 2015) and eosin Y (HARI; KO, 2014; MENG et al., 2013).
These photochemical reactions can be carried out in microstructured reactors. Which has been the theme of several works due to the advantages of these systems, such as homogeneity in the spatial illumination and improvement in light penetration when compared to the large-scale reactors (MATSUSHITA et al., 2007).
2.2.2 Photochemistry in microreactors
The microflow chemistry technology has attracted more attention along the years and became an independent field of research. An advantage of microflow reactors over conventional batch equipment is the superior light penetration in the system. Photochemistry in batch reactors is limited by to the Lambert-Beer law, which states that the light intensity decreases logarithmically along a path due to photon absorption. This limitation is not relevant in microflow systems since the narrow channels impose very low resistance to photon transport (MEYER et al., 2007; WIELAND, 2016).
Moreover, another benefit of carrying out chemical reactions in small volumes is the easier control of reaction parameters, e.g., pressure, temperature, residence
time and flow rate, resulting in an improvement of the conversion and enhanced energy efficiency (MEYER et al., 2007; OELGEMÖLLER, 2012). Photochemical reactions in flow can occur in short time intervals when compared to batch reactors, minimizing by-product formation and increasing the by-productivity of the process (CAMBIÉ et al., 2016).
Given the well-established advantages of microflow systems over batch reactors, several studies have been conducted to evaluate the performance of these devices in photochemical processes (AIDA et al., 2012; AKWI; WATTS, 2018; YOSHIDA; KIM; NAGAKI, 2011). Although commercial microreactors have been widely adopted in photochemistry, several devices built in-house can also be found in the literature, allowing to adjust the characteristics of the microreactor according to the end application (COYLE; OELGEMÖLLER, 2008). Noël (2017) established that the correct selection of the reactor material is of paramount importance, since the majority of the capillaries used in microflow chemistry are made of perfluoroalkoxyalkane (PFA) and perfluoroethylenepropylene (FEP), and these materials tend to present a fast degradation, especially at higher energy wavelengths. Thus, studies have suggested the design of photomicroreactors based on alternative materials, such as quartz, glass, silicon, metal, and ceramic, accordingly to the type of reaction that will be carried out (COYLE; OELGEMÖLLER, 2008).
Another point for improvement is the energy efficiency of the photomicroreactors (NOËL, 2017). Thus, inspired by the Luminescent Solar Concentrator (LSC) technology, Cambié et al. (2017) developed a luminescent solar concentrator-based photomicroreactor (LSC-PM) to promote chemical reaction activated by sunlight. This photomicroreactor is composed by a synergistic integration of LSC and microflow chemistry, which allows an enhancement in the energy efficiency since it harvests the sunlight and waveguides this energy to the microchannels to promote the photochemical reactions (CAMBIÉ et al., 2017a,c; ZHAO et al., 2018b). The LSC-PM will be the theme of this work and, in order to contextualize it, the item 2.3 will give a general approach of the system, and further details will be explored in the next topics.
2.3 LUMINESCENT SOLAR CONCENTRATOR-BASED PHOTOMICROREACTOR (LSC-PM)
The LSC-PM was first presented in 2017 by Cambié et al., with the objective of enabling chemical reactions driven by sunlight. As previously stated, the device merges microflow chemistry and the LSC principle, i.e., the flow reactor was embedded in an LSC light guide, a dye-doped base of polydimethylsiloxane (PDMS), in which the luminescent particles were selected in a way that its emission profile matched the absorption spectrum of the photosensitizer. The dye-doped base of PDMS down-converts the energy to a narrow wavelength region and transports this energy towards the embedded microchannels, where the flowing reactants are transformed (CAMBIÉ et al., 2017a).
To evaluate the performance of the LSC-PM a benchmark reaction was used in the system, the [4+2] cycloaddition of 9,10-diphenylanthracene (DPA) with singlet oxygen that was generated via photosensitization in the presence of methylene blue (MB) (ZHAO et al., 2018a). In order to highlight the components of the LSC-PM, this work was segregated into two categories: light guide and flow. An illustration of the system’s structure can be observed in the scheme presented in Figure 3.
Moreover, the device was built in a way that it could work either with diffuse or direct light. A control system was developed in order to enable constant conversion independent of the sunlight fluctuations during the day. The device and its characteristics will be elucidated in detail in the next topics.
2.3.1 LSC principle
The LSC system was proposed more than thirty years ago by Goetzberger and Greubel (1977). The project aimed an inexpensive and efficient energy generation system, where one of the main characteristics was the operation with direct and diffuse solar radiation, as a consequence of its high acceptance angle for incident light (GOETZBERGER; GREUBEL, 1978, 1977; WIELAND, 2016).
Figure 3 - Schematic representation of the LSC-PM system.
Reference: adapted from Cambié et al. (2017) and Zhao et al. (2018).
The operation of the system is quite simple, consisting of a light trapping and guiding concept. The system collector consists on a flat sheet of glass or polymer (more common), such as PDMS (CHOU; HSU; CHEN, 2015; YANG et al., 2013) and polymethylmethacrylate (PMMA) ) (LIU et al., 2014; MEINARDI et al., 2014), with a refractive index n, doped with fluorophores. The fluorophores function is the absorption and reemission of the absorbed light in high quantum efficiency (SARK et al., 2008). Quantum dots, rare earths, and organic dots can be cited as the frequently employed fluorophores.
The quantum yield of the fluorophore will be responsible for the emission of the light in a longer wavelength or the dissipation of the absorbed light to heat. When emitted, the photon experiences a wavelength shift and, since the emission reaches the outside of the escape cone, the light is trapped through total internal reflection (TIR) in the LSC matrix (TUMMELTSHAMMER, 2016). Thus, the trapped radiation is
wave-guided and reaches the edges of the concentrator, where it can be converted into electricity by photovoltaic cells (Figure 4).
Figure 4 - Representation of the LSC system.
Reference: adapted from Richards (2006).
However, the LSC presents energy losses in the system. As Figure 4 demonstrates, reflection, re-absorption, and scattering can be considered some of the “photon fates” in the system, characterizing energy losses, since the ideal condition would be the TIR to the PV cell. This theme will be discussed in detail in this work.
Therefore, giving the broad application of the LSCs, different approaches have been reported aiming to improve their performance, such as the minimization of luminophore self-absorption losses (ERICKSON et al., 2014; KRUMER et al., 2013, 2017; SUMNER et al., 2017; WU et al., 2010), enhancement of power conversion efficiency (CORRADO et al., 2013; DAS; NARAYAN, 2013; DESMET et al., 2012; SARK et al., 2008), and improvement of photon transport (ILAN; KELLEY, 2011; RONCALI; GARNIER, 1984).
2.3.2 Host material
The LSC-PM matrix is inspired in the LSC one. Therefore, a fundamental factor in an LSC is the selection of the light guide. Ideally, the host material must be inexpensive, highly transparent, with a broad refractive index (~1.5), photo-stable, and mechanically and chemically resistant. Thus, polymers have been frequently chosen as light guides since several materials in this class meet the required characteristics (EBRAHIMIPOUR; ASKARI; RAMEZANI, 2016).
Among the several options available, PMDS and PMMA are the most common choice. Despite the several studies involving PMMA (CORRADO et al., 2013; KERROUCHE et al., 2014; KRUMER et al., 2013; LIU et al., 2014; VAN SARK, 2013), Chou, Hsu and Chen (2015) alleged that this rigid substrate can restrict the applicability of the technology, and involves complicated fabrication techniques. Thus, PDMS became an interesting alternative for the construction of LSC devices, being studied in several works (APAKONSTANTINOU, 2016; CHOU; CHUANG; CHEN, 2013; MEINARDI et al., 2014).
In fact, PDMS is the favorite host material for microscale fluid devices. This fact can be explained by its advantages, such as the low production cost compared to substrate materials such as silicon or glass, which allow fast prototyping; optical transparency (above wavelengths ~230 nm) and flexibility (BHAGAT; JOTHIMUTHU; PAPAUTSKY, 2007). Moreover, PDMS can be easily doped with Lumogen F Red 305 (reference dye) (HOFMANN et al., 2006). Besides the criteria already cited, the refractive index, 𝑛, is also an important parameter for the design and optimization of these devices. The refractive index is, briefly, the ratio of the velocity of light of a certain wavelength in the vacuum relative to the velocity of light passing through a specific material (BRYDSON, 2017). The refractive index interferes directly in the proportion of the trapping-guiding process of the photons inside the device and, according to Mouedden (2016), the higher the refractive index, the larger will be the number of trapped photons. PDMS presents a moderate refractive index of 1.41 (CAMBIÉ et al., 2017a).
2.3.3 Luminescent particles
The doping of the material is pivotal to the system since the luminescent particles are responsible for the absorption and reemission of the radiation in a longer wavelength. As already stated in item 2.3.1, the commonly utilized luminescent particles in LSC are organic dyes, quantum dots, and rare earths materials. Among these materials, organic dyes have been the most commonly used, due to their low cost, easy production, and abundance.
Among the species of organic dyes, since the 1970s and 1980s, the most used ones for this purpose are coumarins, rhodamines, and perylenes derivatives (MANZANO CHÁVEZ, 2017). The Lumogen F Red 305 (Red 305), from the perylene group, has been broadly applied in LSC systems (DESMET et al., 2012; DIENEL et al., 2010; KRUMER et al., 2017; SLOOFF et al., 2008) due its good solubility, high quantum yields and broad Stokes shift (BALABAN, 2013). Furthermore, this compound presents higher photostability than the others Lumogen F group dyes (BASF, 1997).
2.3.3.1 Light transport and loss factor
In the LSC-PM the ideal condition for light transport is the TIR, where the luminescent particle absorbs the light and re-emit it in a way that a TIR in the system guides it to the reaction centers. However, as already stated, the LSC system and consequently, the LSC-PM, still presents limited efficiencies due to energy losses in the system. These losses can be classified from the observation of the possible photon fates, and those photon fates can be seen in Figure 5.
Figure 5 - Representation of possible fates of the photons in the moment that the light strikes the LSC-PM device.
Reference: Cambié et al. (2017).
As shown in Figure 5 the photon fates in the system can be classified as:
• reflection (1) and transmission (2): According to Chavéz (2017), when light impinges on a flat surface it passes through a medium with a refractive index 𝑛1 to a medium with a different refractive index, 𝑛2. This can result in two fates: reflection and transmission through the material. These characteristics determine a Fresnel loss type material, where part of the light is reflected and part of the light is transmitted.
• emission (3): this loss can be induced by the incidence of light in the scape cone, where it can cause a top or bottom emission.
• edge emission (4): in the LSC-PM system, the edge emission is not desired and, consequently, it is considered a loss.
• reaction media absorption (5): occurs when light is absorbed by the media and not re-emitted to the system.
• non-radiative losses (6): a non-radiative dissipation of the energy previously absorbed by the luminescent particles.
In order to optimize the performance of the light-dependent systems and quantify these losses, ray-tracing simulations are of paramount importance. This can be considered a result of the ability of these algorithms to elucidate the photon behavior in the analyzed domain.
2.3.3.1.1 Ray-tracing simulations
Currently, ray-tracing simulation is an approach broadly applied in several areas such as radio (ATHANAILEAS et al., 2010; CHEN; DELIS; BERTONI, 2004), acoustics (JANG; HOPKINS, 2018; MO et al., 2016), gravitational waves (DING; WAN; YUAN, 2003; JONES; BEDARD, 2018; VADAS; FRITTS, 2009) and optics (HU et al., 2015; KERROUCHE et al., 2014; REHMAN, 2019).
Trough optical ray-tracing simulation, it is possible to simulate the path of the rays by combing the principles of traditional geometric optics and the Monte Carlo method. Dissipation of light is determined by the optical properties of surfaces, materials and emission sources (KERROUCHE et al., 2014). Since these simulations work through the stochastic Monte-Carlo probability method, random numbers are considered to determine the direction, position, and energy, among other parameters (JI; ZHANG, 2019). The fate of the rays can be elucidated since these algorithms are able to evaluate the rays that are absorbed, reflected, refracted, diffracted and scattered in the domain.
Thus, the application of ray-trace modeling can be considered, nowadays, indispensable for the design, performance evaluation and optimization of optical systems. In this context, several studies have been devoted to applying ray-trace modeling in those systems. Chávez (2017), simulated three LSC models, the first one made of a PMMA matrix and doped with a red dye. The second and third LSC models were doped with thulium: one of them was a glass doped with thulium particles, while the other one consisted of a pure glass coated with thulium. The main objective was to
find the characteristics that affect the optical efficiency of the system and optimize these parameters. The simulations were carried out using the optical engineering software LightTools®. The optical losses in the red dye model were elucidated and quantified and then compared with experimental data, showing a good agreement and enabling the validation of the model. On the thulium models, the author concludes from the results of the simulation that the LSC performance might be improved by adding thin film layers in the glass, such as anti-reflection coating and selective filters to allow visible light and reflect near-infrared photons into the glass.
Haines et al. (2012), evaluate the effect of the perylene diimide in PMMA films cast onto glass substrates. The Monte Carlo ray-trace algorithm was developed using MATLAB® software. The ray-tracing results were compared with the experimental ones and showed a good agreement. From the data obtained the authors concluded that the major factor that affects the performance of the LSC is the luminescence quantum yield of the dye. Moreover, with the simulation data coupled with the experimental output, the authors concluded that the perylene diimide is not an optimized material for LSC applications.
Kumar, Velu, and Balasubramanian (2019) proposed a novel free form lens design for collimating UV light from an LED, where four models were proposed: three surfaces of free form and one with total internal reflection. A ray-tracing simulation with the ZEMAX OpticsStudio® 15.5 ray-tracing software was employed to observe the performance of the lens. The authors concluded from the ray-tracing simulations that the designed lens presented a better collimation angle of ±2° with improved efficiency of 58.88% when compared to 34% of the existing lens.
Vishwanathan et al. (2015) compared the performance of the flat and bent LSC-photovoltaic (LSC-PV) doped with Lumogen Red 305 dye, via Monte Carlo ray-tracing simulations on LightTools® software. The authors modeled two types of light sources in order to simulate the direct and diffuse irradiance, the first one was a perpendicularly directed light source and the second one a Lambertian source. The sun sources were modeled with the AM 1.5G spectrum. The best optical efficiencies were found for the flat LSC-PV, an observation indicated as a consequence of the higher loss of the rays in the bent LSC, due to its bent light guide. The concentration
of the dye was varied in the simulations as well. From the simulations, the authors observed that the low concentration of dye resulted in insufficient absorption of the light. However, higher concentration leads to losses by reabsorption, decreasing its optical efficiency. The efficiency of the system with very low concentrations of the dye was about 10% and increasing the dye concentration resulted in a maximum efficiency of 20%. This maximum efficiency can be explained by the excessive reabsorption and eventually generation of non-emissive dye clusters in the light guide.
Cambié et al. (2017) simulated the LSC-PM photon path with a modified version of PvTrace, a Python-based Monte-Carlo ray-tracing algorithm for LSC simulations, in order to guide further design and optimization of the device. Good agreement was found between the experimental and simulated data. Based on the simulations, the authors were able to optimize the system from an exchange in the dimensions of the channels.
Therefore, it is evident from the studies reported above that there is a clear need to elucidate the losses in the optical systems in order to evaluate their performance and optimize them. This scenario motivated the studies carried out in this work.
2.3.4 Flow characterization
The oxygen singlet photo-oxidation of 9,10 diphenylanthracene (DPA) to its corresponding endoperoxide has been extensively studied since it is a light-dependent reaction, allowing simple quantification of the reaction advance (PITRE et al., 2015).
The molecular oxygen became an interesting “green reagent” due to its negligible environmental impact. Moreover, other advantages are its low cost and large availability (SEEBERGER, 2011). There are many ways to generate oxygen singlet, including the photosensitized generation. It can be explained due to the fact that the photosensitized generation is a simple and controllable method, requiring only light in an appropriate wavelength, oxygen, and a photosensitizer able to absorb this energy and excite the oxygen to its singlet state (DEROSA; CRUTCHLEY, 2002). The
sensitizing potential of rose bengal, methylene blue, and hematoporphyrins can be highlighted (KRUK, 1998).
This oxidation can occur in several ways, where the ene, [2+2] and [4+2] cycloadditions can be highlighted as powerful methods for the addition of molecular oxygen (CLENNAN; PACE, 2005). The [4+2] mechanism for the reversible binding of 1O2 in aromatic compounds is widely exploited for the production of chemical traps, having in mind that the endoperoxide formed can be considered a specific product for the reaction with 1O2. The detection of these endoperoxides formed in the reaction can be provided through a HPLC–MS analysis (MARTINEZ et al., 2006).
Therefore, for the LSC-PM system, a [4+2] cycloaddition of singlet oxygen generated via methylene blue photosensitization to 9,10 diphenylanthracene was used as a benchmark reaction since a light-dependent reaction can be monitored through UV-Vis spectrophotometry (Figure 6). Acetonitrile can be used as a solvent due to its non-absorbing property (CAMBIÉ et al., 2017a).
Figure 6 - 9,10 diphenylanthracene conversion to its respective endoperoxide.
Reference: Cambié et al. (2017).
Despite the several advantages of the use of sunlight already stated, the necessity of constant irradiation for the chemical synthesis represents a challenge, due to the constant fluctuations of photon flux during the day, i.e., weather changes are of great impact on the reaction efficiency. Thus, a control system to mitigate the impact of the fluctuation is crucial.
2.3.4.1 Control system
The constant evolution of methods for chemical synthesis has been closely accompanied by a higher degree of automation of the equipment involved. In this context, low-cost digital control can be successfully applied to the improvement of the performance of chemical reactors applied to the synthesis of added-value compounds (FITZPATRICK; BATTILOCCHIO; LEY, 2016).
Photochemical synthesis based on sunlight is one of the fields that clearly presents a need for control systems. The changes in the irradiation during the day has a huge impact on the efficiency of the reaction. Although intensive research has been conducted for the automatization of these systems, on-line analytical techniques still represent the strategy most applied (FABRY; SUGIONO; RUEPING, 2014). However, this approach is not indicated for processes driven by sunlight, once the light irradiation can change drastically and high time delays in the measurements would generate undesirable conversion fluctuations (ZHAO et al., 2018).
Based on this scenario, Zhao et al. (2018), proposed a real-time reaction control system for the LSC-PM, in order to maintain a constant conversion despite the irradiation fluctuations during the day. This control system works through the adjustment of the reactor’s residence time based on the light intensity in a specific time of the day.
Figure 7 - Representation of the LSC-PM real-time reaction control system.
Reference: Zhao et al. (2018).
Figure 7 shows a representation of the items that constitute the real-time reaction control system. A light sensor is placed at the edge of the microreactor and is responsible for monitoring the photon flux in the device. Since the light intensity at the LSC-PM’s edge is proportional to the photon flux reaching the microchannels, this measurement can be used to characterize the optical performance of the device. This light sensor is connected to a microcontroller that automatically adjusts the pump power depending on the light intensity read. The time delay involved in reading the voltage, executing the control algorithm and sending the corresponding action to the pump is c.a. 500 ms. As previously indicated, the reaction of [4+2] cycloaddition of singlet oxygen to 9,10 diphenylanthracene can be continuously analyzed via UV-VIS spectrometer, since the reaction kinetics is light-limited (ZHAO et al., 2018).
Thus, it is possible to declare that the light distribution is not the only parameter to be considered when analyzing and optimizing the LSC-PM applied to the synthesis of added-value compounds: the microflow also has a huge effect on the reactor’s performance. Characterizing the flow is pivotal for the optimization of the microreactor and computational fluid dynamics (CFD) algorithms can be successfully applied for this purpose (CHETVERUSHKIN et al., 2004).
2.3.5 CFD applied to flow distribution characterization
CFD is a powerful tool that can be applied in several engineering fields. This tool can be employed to observe the physical events that occur in the flow of fluids on a determinate domain. These events can be often related to the phenomena associated with dissipation, diffusion, convection, boundary layers, and turbulence. CFD works with fluids in motion and analyzes how the behavior of the fluid can influence processes such as heat and mass transfer. It works through the use of algorithms and numerical methods to provide flow information that with traditional techniques would be considered difficult, expensive or inconceivable, e.g., analytical solutions or experiments (TU; YEOH; LIU, 2018).
Thus, in this context, CFD has been shown to be useful for the analysis of the flow characteristics and reactor performance, being able to capture the presence of dead zones, the velocity and concentration fields, and evaluate the transport of radiation. In this sense, several studies have been carried out to investigate the flow behavior in photomicroreactor. Investigations on the design and optimization of microreactors with these tools have been highly successful for liquid phase systems. This can be attributed to the fact that the behavior of incompressible liquid along with the laminar flow regime, characteristic of micrometric scale, usually generate good results on CFD simulations (SANTANA; SILVA; TARANTO, 2019). Therefore, many studies have applied CFD tools to observe the significant influences of microfluidic geometries and flow on the performance of the operating system (COMMENGE et al., 2002; GRIFFINI; ASTERIOS, 2007; O-CHAROEN; SRIVANNAVIT; GULARI, 2007; ODIBA et al., 2016).
Chetverushkin et al. (2004) showed the possibility to integrate the reactor hydrodynamics, radiation intensity, and kinetics. The model was applied to several types of reactors. The authors conclude that the difference in the performance of the can be due to the light intensity distribution in the reactors, which changed when the lamp was repositioned. Moreover, the authors conclude that the UV transmittance of the fluid and the fluid flow rate also interferes on the reactor performance.
Based on that, it is evident the range of possible uses of CFD for accurate modeling of fluid flow in photochemical reactors, which motivated part of the studies performed in this work.
3 MATERIALS AND METHODS 3.1 CFD SIMULATION
The CFD modeling was based on the indoor kinetic investigation carried out by Zhao et al. (2018). The steps of the investigation can be seen in Figure 8. The syringe pump is responsible to insert the solutions with DPA and MB (0.2 mM and 0.4 mM in acetonitrile respectively to the microreactor). The white led strip presented in Figure 8 is placed onto a cylindrical box, and a power supply is responsible to power this led strip. In the outlet of the microreactor the reaction mixture is analyzed by a UV-VIS spectrometer, where the absorbance is further utilized to calculate the conversion of the reaction. In order to observe the behavior of the system in different light intensities the LED strip intensity was varied. The kinetic curves were obtained through a flow rate change under each light intensity, and the correspondent voltage value under each light intensity detected by the microcontroller was recorded as well (Zhao et al. 2018). The data utilized in this work was provided by the Noël Research Group and can be seen in Appendix A, Table A1.
Figure 8 – Flow scheme of the indoor kinetic experimental procedure.
3.1.1 Geometry
The geometry was built in three dimensions to simulate laminar and single-phase flow, as well as the transport of chemical species. All the sections of the experimental device were taken into account to enable an optical analysis, which will be described in a further section. Figure9shows the microreactor geometry assembled in COMSOL Multiphysics® (version 5.3a) software and the dimensions of it. The model was built based on the device proposed by Cambié et al. (2018), and it consists of an inlet (A), a reaction channel (B), and a mold border (C).
Figure 9 – (1) Representation of the LSC-PM built in COMSOL Multiphysics® (version 5.3a) software. (2) Dimensions of the LSC-PM device.
Reference: author (2019).
3.1.2 Grid independence test
In order to evaluate the degree of numerical uncertainty due to the construction of the mesh, the method proposed by Celik et al. (2008), the so-called Grid Convergence Index (GCI), was applied. The conversion of DPA at the outlet of the photomicroreactor was used for the implementation of the method. To calculate the
